Stratum deformation early warning method and device based on multi-parameter dynamic monitoring

Abstract

The invention belongs to the field of geological environment monitoring, particularly relates to a stratum deformation early warning method and device based on multi-parameter dynamic monitoring, and aims to solve the problems of false alarm and missing alarm in a complex construction environment caused by unreasonable weight distribution and rigid early warning mechanism. The method comprises the steps of obtaining an original data set; calculating the information amount between the standard deviation of each parameter and the correlation coefficient and the weight values of the plurality of parameters, weighting the parameters according to the weight values, and constructing a to-be-trained data set; inputting the to-be-trained data set into a pre-constructed stratum deformation prediction model for training; acquiring data in real time according to a set acquisition frequency as input data, and inputting the input data into the trained stratum deformation prediction model to obtain stratum deformation prediction data; and executing a corresponding disposal plan according to the early warning level corresponding to the stratum deformation prediction data. According to the invention, the weight distribution is more reasonable, the precision and stability of stratum deformation prediction are improved, and the early warning is more accurate.

Description

Background Art

[0002] In many engineering construction projects, such as urban subway construction, high-rise building piling, and land reclamation projects in coastal areas, changes in geological parameters such as ground subsidence, pore water pressure changes, soil pressure and groundwater level changes have a key impact on project stability and surrounding environmental safety.

[0003] Land subsidence is usually caused by over-extraction of deep confined water. Excessive exploitation of confined water will cause the groundwater level to drop, the pore water pressure to decrease, and thus increase the effective stress. Under the condition of increasing effective stress, the compressible strata will be compressed, and eventually form subsidence on the surface. Therefore, the hydrodynamic conditions related to the groundwater level and pore water pressure and the stratum characteristics related to soil pressure are the most direct factors affecting land subsidence.

[0004] Traditional monitoring methods for ground subsidence, groundwater level and pore water pressure usually require the establishment of different monitoring equipment, such as ground subsidence monitoring stations, soil pressure pore water pressure sensors and water level sensors, etc., and the measurement of various parameters needs to be carried out separately. Traditional monitoring methods mostly rely on single data, and it is difficult to comprehensively consider the synergistic effects of multiple factors such as pore water pressure, groundwater level changes, and soil pressure on stratum deformation, resulting in limited prediction accuracy and a lack of intelligent error warning mechanism, which makes it impossible to detect potential geological risks in a timely manner.

[0005] Currently, when using multiple factors such as pore water pressure, groundwater level, and soil pressure to predict stratum deformation, there are problems such as unreasonable weight distribution and rigid early warning mechanism, which often lead to false alarms and missed alarms and cannot meet the needs of complex and changeable construction environments.

[0006] Based on this, the present invention proposes a formation deformation early warning method and device based on multi-parameter dynamic monitoring. Summary of the invention

[0007] In order to solve the above-mentioned problems in the prior art, that is, currently, when using multiple factors such as pore water pressure, groundwater level, soil pressure, etc. to predict stratum deformation, there are problems such as unreasonable weight distribution and rigid early warning mechanism, which often lead to false alarms and missed alarms, and cannot meet the needs of complex and changeable construction environments. The present invention provides a stratum deformation early warning method and device based on multi-parameter dynamic monitoring.

[0008] In a first aspect, the present invention provides a stratum deformation early warning method based on multi-parameter dynamic monitoring, the method comprising: Acquire multiple parameters for monitoring stratum deformation and time series data of stratum deformation data as original data sets, wherein the multiple parameters at least include pore water pressure, groundwater level change and soil pressure; Eliminate outliers in the original data set and perform normalization processing; Calculating the standard deviation of multiple parameters and the correlation coefficient with formation deformation respectively, and calculating the information amount between the standard deviation and the correlation coefficient of each parameter, calculating the weight values ​​of multiple parameters based on the information amount, weighting the normalized multiple parameters according to the weight values, obtaining the input feature vector after weighting, and constructing the data set to be trained; Inputting the data set to be trained into a pre-built stratum deformation prediction model for training; wherein the stratum deformation prediction model is built based on a GAN-LSTM network; The pore water pressure, groundwater level change and soil pressure of the formation are collected in real time according to the set collection frequency as input data, and input into the trained formation deformation prediction model to obtain formation deformation prediction data; The absolute error between the formation deformation prediction data and the historical formation deformation data is calculated in real time to form an error sequence. The warning threshold is calculated based on the error sequence. The warning threshold is adjusted according to the deformation trend of the formation deformation prediction data within a set time. The warning level is divided according to the adjusted warning threshold. According to the warning level corresponding to the formation deformation prediction data, the corresponding disposal plan is executed.

[0009] Furthermore, the data set to be trained is input into a pre-built formation deformation prediction model for training, and the training method is as follows: The data set to be trained is used as input data and is sequentially input into the formation deformation prediction model according to time steps. The input data is analyzed in time series through the LSTM network. Use a fully connected layer to extract features from the hidden state of the last time step of the LSTM network to predict the formation deformation value; The mean square error is used as the loss function to adjust the parameters of the LSTM network, and the loss function is minimized by the Adam optimization algorithm; The trained LSTM network is used as the generator of the GAN network to generate formation deformation prediction data; Construct a multi-layer perceptron as the discriminator of the GAN network to distinguish the real stratum deformation data from the stratum deformation prediction data output by the generator; Adjust the parameters of the generator and discriminator based on a preset loss function; The Adam optimizer is used alternately to update the generator and discriminator parameters, and a set number of rounds of iterative training are performed until convergence.

[0010] Furthermore, the warning threshold is calculated based on the error sequence, and the method is as follows: ; in, is the first level warning threshold, is the second level warning threshold, is the mean of the error sequence, is the standard deviation of the error series, , is the preset empirical coefficient.

[0011] Furthermore, the warning threshold is adjusted by the deformation trend of the formation deformation prediction data within a set time, and the method is as follows: Based on the sliding window algorithm, the past k Linear fitting is performed on the formation deformation prediction data of each time step and the slope is obtained as the deformation trend; If the slope is positive and greater than a preset threshold, lower the warning threshold; If the slope is negative, the warning threshold is increased.

[0012] Further, the warning threshold is lowered and the warning threshold is increased by: Lower the warning threshold: ; Increase the warning threshold: ; in, To lower or raise the warning threshold, i is 1 or 2, is the original warning threshold, is the adjustment factor.

[0013] Furthermore, the warning levels are divided according to the adjusted warning thresholds, and the method is as follows: When the error When there is no warning; when It is a Level 1 warning when It is a level 2 warning.

[0014] In a second aspect of the present invention, a stratum deformation early warning device based on multi-parameter dynamic monitoring is proposed, based on a stratum deformation early warning method based on multi-parameter dynamic monitoring, the device comprises: a fixed pile, a groundwater level monitoring component, a pore water pressure monitoring component, a soil pressure monitoring component and a ground settlement monitoring component; The fixing piles are fixedly implanted in the foundation and the top ends thereof extend out of the ground to a set height, so as to support the components; The underground water level monitoring assembly includes a water level sensor buried in the foundation and arranged in the water level layer to monitor the change of the underground water level; The pore water pressure monitoring assembly includes pore water pressure gauges buried in soil layers at different depths, and used to measure the pore water pressure in each soil layer in real time; The soil pressure monitoring component includes a soil pressure sensor buried in the rock and soil layer, which is used to measure the soil pressure in real time; The ground settlement monitoring assembly includes a reference plate fixed on the ground and a displacement sensor. The displacement sensor is connected to the reference plate and is arranged under the reference plate and the fixed piles to monitor the displacement of the reference plate and obtain stratum deformation data.

[0015] Furthermore, the groundwater level monitoring component also includes a groundwater level monitoring pipe, a pipe wall of which is provided with a water-permeable hole, a filter is installed on the water-permeable hole, and a water level sensor is installed inside the groundwater level monitoring pipe.

[0016] Furthermore, the device also includes a data acquisition and storage unit and a data transmission and early warning unit. The data acquisition and storage unit is electrically connected to the water level sensor, the pore water pressure gauge, the soil pressure sensor and the displacement sensor, and is used to obtain pore water pressure, groundwater level changes, soil pressure and ground subsidence data, and transmit them to the remote monitoring center through the data transmission and early warning unit.

[0017] Furthermore, the device also includes a power supply module, which includes a solar panel and a lithium battery pack. The solar panel is used to charge the lithium battery pack. The lithium battery pack and the solar panel are both electrically connected to the water level sensor, the pore water pressure gauge, the soil pressure sensor and the displacement sensor and provide power.

[0018] Beneficial effects of the present invention: High integration: Integrating ground subsidence, pore water pressure and water level monitoring functions into one device, reducing the number of equipment and installation complexity, and lowering monitoring costs.

[0019] Convenient data collection: Remote data transmission is achieved through the wireless communication module, eliminating the need for manual on-site collection, which improves the efficiency and accuracy of data collection.

[0020] Multi-factor weight assignment and dynamic update: Weights are assigned to the three factors of pore water pressure, groundwater level and soil pressure, and the weights are updated dynamically and regularly, taking full account of the differences and correlations between the factors, making the weight assignment more reasonable and improving the accuracy and reliability of the prediction model.

[0021] GAN-LSTM model optimization: GAN is combined with LSTM, and GAN is used to generate false data to enhance the robustness of the model. LSTM is then used for secondary learning to mine time series features, effectively improving the accuracy and stability of stratum deformation prediction.

[0022] Multi-level warning mechanism: multi-level warning is divided according to the error size, deformation trend, etc., corresponding to the disposal plans of different emergency levels, making the warning more practical and targeted, providing sufficient time for the prevention and control of geological disasters, and has important practical application value. BRIEF DESCRIPTION OF THE DRAWINGS Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings: Figure 1 It is a schematic flow chart of a stratum deformation early warning method based on multi-parameter dynamic monitoring of the present invention; Figure 2 It is a structural schematic diagram of a formation deformation prediction model in a formation deformation early warning method based on multi-parameter dynamic monitoring of the present invention; Figure 3 It is a structural schematic diagram of a stratum deformation early warning device based on multi-parameter dynamic monitoring of the present invention. DETAILED DESCRIPTION

[0024] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0025] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0026] The first embodiment of the present invention provides a stratum deformation early warning method based on multi-parameter dynamic monitoring, the method comprising: Acquire multiple parameters for monitoring stratum deformation and time series data of stratum deformation data as original data sets, wherein the multiple parameters at least include pore water pressure, groundwater level change and soil pressure; Eliminate outliers in the original data set and perform normalization processing; Calculating the standard deviation of multiple parameters and the correlation coefficient with formation deformation respectively, and calculating the information amount between the standard deviation and the correlation coefficient of each parameter, calculating the weight values ​​of multiple parameters based on the information amount, weighting the normalized multiple parameters according to the weight values, obtaining the input feature vector after weighting, and constructing the data set to be trained; Inputting the data set to be trained into a pre-built stratum deformation prediction model for training; wherein the stratum deformation prediction model is built based on a GAN-LSTM network; The pore water pressure, groundwater level change and soil pressure of the formation are collected in real time according to the set collection frequency as input data, and input into the trained formation deformation prediction model to obtain formation deformation prediction data; The absolute error between the formation deformation prediction data and the historical formation deformation data is calculated in real time to form an error sequence. The warning threshold is calculated based on the error sequence. The warning threshold is adjusted according to the deformation trend of the formation deformation prediction data within a set time. The warning level is divided according to the adjusted warning threshold. According to the warning level corresponding to the formation deformation prediction data, the corresponding disposal plan is executed.

[0027] In order to more clearly explain the stratum deformation early warning method based on multi-parameter dynamic monitoring of the present invention, the following is combined with Figure 1 The embodiments of the present invention are described in detail as follows: Acquire multiple parameters for monitoring stratum deformation and time series data of stratum deformation data as original data sets, wherein the multiple parameters at least include pore water pressure, groundwater level change and soil pressure; In this embodiment, pore water pressure ( P ), groundwater level changes ( H )、Soil pressure( S ) and stratum deformation ( D ) to form the original data set , where N is the number of samples.

[0028] Eliminate outliers in the original data set and perform normalization processing; In this embodiment, in the data processing center, statistical methods, such as the three-times standard deviation rule, are used to identify and remove outliers in the pore water pressure, groundwater level, soil pressure, and stratum deformation data to prevent them from interfering with subsequent model training.

[0029] Normalization: The minimum-maximum normalization algorithm is used to uniformly map the data collected by the four types of sensors to the [0, 1] interval. The formula is as follows: For pore water pressure: ; For groundwater level changes: ; For soil pressure changes: ; For stratum deformation changes: .

[0030] Calculating the standard deviation of multiple parameters and the correlation coefficient with formation deformation respectively, and calculating the information amount between the standard deviation and the correlation coefficient of each parameter, calculating the weight values ​​of multiple parameters based on the information amount, weighting the normalized multiple parameters according to the weight values, obtaining the input feature vector after weighting, and constructing the data set to be trained; The standard deviation of multiple parameters is calculated as: Standard deviation of pore water pressure ,in is the mean of the normalized pore water pressure data, similar to the standard deviation of the groundwater level. and the standard deviation of soil pressure , I will not go into details here.

[0031] The correlation coefficient between each parameter and the formation deformation is specifically calculated by calculating the Pearson correlation coefficient with the formation deformation. The method is: Correlation coefficient between pore water pressure and formation deformation : ; in, is the normalized mean value of the formation deformation data.

[0032] Similarly, the correlation coefficient between groundwater level and stratum deformation is calculated and the correlation coefficient between soil pressure and stratum deformation .

[0033] The amount of information is calculated as: Information content of pore water pressure Similarly, the information of groundwater level is obtained and soil pressure information .

[0034] The weight values ​​of multiple parameters are calculated based on the information amount, and the method is as follows: Pore ​​water pressure weight , the weight value of the groundwater level can be calculated in the same way and the weight of soil pressure .

[0035] The normalized pore water pressure, groundwater level change, and soil pressure data are weighted to obtain a new input feature vector: , build a new dataset .

[0036] Inputting the data set to be trained into a pre-built stratum deformation prediction model for training; wherein the stratum deformation prediction model is built based on a GAN-LSTM network; In this embodiment, the data set to be trained is input into a pre-built formation deformation prediction model for training, and the training method is as follows: The data set to be trained is used as input data and is sequentially input into the formation deformation prediction model according to time steps. The input data is analyzed in time series through the LSTM network. See also Figure 2 , the input dimension is 3 (the dimension of the weighted feature vector), and the hidden layer dimension is set to ,Include Hidden layers. Input data , by time step Input sequentially, the internal state update formula of LSTM unit is as follows: Input Gate: ; Forget Gate: ; Output Gate: ; Candidate memory gates: ; Memory Unit: ; Hidden state: ,here is the weight matrix, is the bias vector, is the sigmoid function.

[0037] Use a fully connected layer to extract features from the hidden state of the last time step of the LSTM network to predict the formation deformation value; The hidden state of the last time step of LSTM Access the fully connected layer and output the predicted ground deformation value ,Right now .

[0038] The mean square error is used as the loss function to adjust the parameters of the LSTM network, and the loss function is minimized by the Adam optimization algorithm; Among them, the mean square error is used as the loss function: ; The trained LSTM network is used as the generator of the GAN network to generate formation deformation prediction data; Construct a multi-layer perceptron as the discriminator of the GAN network to distinguish the real stratum deformation data from the stratum deformation prediction data output by the generator; Construct a multi-layer perceptron as a discriminator, with the input being real stratum deformation data And the predicted data output by the generator , output the discrimination probability. For example, the discriminator has two hidden layers, and the input is transformed through the fully connected layer: , and then through the second layer of full connection: ,in is the input data, is the discriminant probability.

[0039] Adjust the parameters of the generator and discriminator based on a preset loss function; Generator loss: , the goal is to make the discriminator recognize the generated data as real data.

[0040] Discriminator loss: ,The goal is to correctly distinguish between real and generated data.

[0041] The Adam optimizer is used alternately to update the generator and discriminator parameters, and a set number of rounds of iterative training are performed until convergence.

[0042] Specifically, the Adam optimizer is used alternately to update the parameters of the generator and the discriminator. Through rounds of iterative training, LSTM can learn twice under the GAN framework to improve prediction performance.

[0043] The LSTM network of the present invention includes an input layer, a convolutional layer, a pooling layer, an LSTM layer and an output layer connected in sequence, which are described in detail as follows: Input Layer Function: Receive input data and provide data basis for subsequent processing modules. Connection method: Connect to the convolution layer and pass the input data to the convolution layer for processing. Features: The dimension and format of the input data need to match the input requirements of the convolution layer. For example, when processing image data, the dimensions of the input layer are usually (number of samples, height, width, number of channels).

[0044] Convolutional Layer Function: Extract spatial features from input data and capture local features and patterns through convolution operations. Connection method: Connected to the input layer to receive input data; connected to the pooling layer to pass the feature map after convolution operation to the pooling layer. Features: Use convolution kernels (filters) to slide on the input data, perform convolution operations, and generate feature maps (activation maps). Parameters such as the size, number, and step size of the convolution kernel can be adjusted according to specific tasks to meet different feature extraction requirements. Activation functions (such as ReLU) are usually used to perform nonlinear transformations on the results of convolution operations to enhance the expressiveness of the model.

[0045] Pooling Layer Function: Downsample the feature map output by the convolution layer to reduce the amount of data and computational complexity while retaining important features. Connection method: Connected to the convolution layer to receive the feature map; connected to the flattening layer to pass the downsampled feature map to the flattening layer. Features: Commonly used pooling operations include MaxPooling and AveragePooling, which take the maximum value and average value in the pooling window as output respectively. Parameters such as the size and step size of the pooling window can be set as needed to control the degree of compression of the feature map.

[0046] LSTM layer (LSTMLayer) function: process time series data, capture temporal dynamic features, and solve long-term and short-term dependency problems. Connection method: connected to the flattening layer to receive the flattened vector; connected to the fully connected layer to pass the output of the LSTM unit to the fully connected layer. Features: The LSTM unit contains structures such as input gate, forget gate, cell state, and output gate, which can dynamically update and store information based on the contextual information of the time series. Through the LSTM layer, the model can effectively learn the time series features in the input data, such as the time dependency of pore water pressure and groundwater level changes. The number of hidden layer units in the LSTM layer can be adjusted according to the complexity of the task and the amount of data to achieve better prediction results.

[0047] Output Layer Function: Output the final prediction result of the model and perform corresponding processing according to the task type (such as regression or classification). Connection method: Connected to the fully connected layer to receive the processed features. Features: In regression tasks, the output layer usually has only one neuron and outputs continuous values, such as the predicted value of stratum deformation. In classification tasks, the number of neurons in the output layer is equal to the number of categories, and the softmax activation function is usually used to convert the output into a probability distribution to represent the predicted probability of each category. Through the combination and connection of the above modules, the GAN-LSTM model can effectively process the spatial and temporal features in the input data, and achieve accurate prediction of complex time series problems such as stratum deformation.

[0048] The absolute error between the formation deformation prediction data and the historical formation deformation data is calculated in real time to form an error sequence. The warning threshold is calculated based on the error sequence. The warning threshold is adjusted according to the deformation trend of the formation deformation prediction data within a set time. The warning level is divided according to the adjusted warning threshold. According to the warning level corresponding to the formation deformation prediction data, the corresponding disposal plan is executed.

[0049] This embodiment collects historical data of stratum deformation over a long span of time, and combines it with real-time monitored pore water pressure, groundwater level changes, soil pressure, and stratum deformation data. These data are sorted in time series to form a comprehensive data set.

[0050] For each predicted time point, calculate the predicted formation deformation value Compared with the actual measured value The absolute error , record the error sequence .

[0051] Based on the error sequence, two methods are provided in this embodiment to calculate the warning threshold. The first method is to calculate the mean of the error sequence regularly (for example, every n new data points, n=50). and standard deviation ; ; in, is the first level warning threshold, is the second level warning threshold, is the mean of the error sequence, is the standard deviation of the error series, , The preset empirical coefficient The present invention uses mean and standard deviation to set the threshold value based on classical statistical theory. This method is suitable for the case where the data is normally distributed. It can well capture the fluctuations in the error sequence and can effectively identify outliers in theory. For technicians, the concepts of mean and standard deviation are simple and clear and easy to understand. This makes the results of this method intuitive and easy to explain to non-professionals. By adjusting , The value of can control the sensitivity of the warning level. For example, in the case of high risk of formation deformation, a stricter threshold (larger α value), while in lower risk situations a more lenient threshold (a smaller α The mean and standard deviation are recalculated after every n new data points, ensuring that the threshold can be dynamically updated with new monitoring data and remain consistent with the actual situation.

[0052] The second type is: Calculate the quantile of the error series, such as The difference between the lower quantile and the upper quantile is scaled and set as the threshold, for example: When new data is added, the quantiles are recalculated and the thresholds are automatically optimized.

[0053] Compared with the mean and standard deviation, the quantile of the present invention is less sensitive to outliers. Therefore, when there are extreme values ​​in the data, the quantile method may provide a more stable and reliable threshold setting. If the error sequence is not normally distributed, but skewed or has other forms, the quantile method is still effective because it does not need to assume the form of the data distribution. Whenever new data is added, the quantile is automatically recalculated and the threshold is adjusted. This mechanism allows the system to automatically optimize the warning threshold based on the latest data without human intervention. By adjusting and Different levels of warning thresholds can be set according to actual needs. For example, an appropriate proportional factor can be selected based on historical data experience to ensure that warnings are neither too frequent nor too late.

[0054] The warning threshold is adjusted by the deformation trend of the formation deformation prediction data within a set time, and the method is as follows: Based on the sliding window algorithm, the past k Linear fitting is performed on the predicted formation deformation data of time steps and the slope is obtained. , as the deformation trend; higher-order features such as second-order derivatives can also be calculated to capture the acceleration or deceleration characteristics of the deformation.

[0055] If the slope is positive and greater than the preset threshold, it indicates that the formation deformation is in an accelerated state. At this time, the warning threshold should be appropriately lowered to make the warning more sensitive. The specific adjustment formula can be: ,in is the original warning threshold, To lower or raise the warning threshold, i is 1 or 2, is an adjustment coefficient, and the adjustment coefficient is preferably 0.1 in this embodiment.

[0056] On the contrary, if the slope If it is continuously negative, it means that the deformation is slowing down. The warning threshold can be appropriately increased to reduce false alarms, such as .

[0057] The method for dividing the warning levels according to the adjusted warning thresholds is as follows: When the error When there is no warning; when It is a Level 1 warning when It is a level 2 warning.

[0058] This method of the present invention can adjust the warning threshold in real time according to the latest formation deformation trend. When accelerated deformation is monitored, lowering the warning threshold allows the system to respond to potential risks more quickly, improving the sensitivity and timeliness of the warning; and when the deformation slows down, appropriately raising the warning threshold can avoid excessive false alarms and make the warning more accurate.

[0059] Using a sliding window algorithm combined with linear fitting not only allows the deformation trend (slope) to be obtained, but also higher-order features such as second-order derivatives can be calculated to capture the acceleration or deceleration characteristics of deformation. This helps to more fully understand the formation behavior and provide richer information for decision-making.

[0060] By dynamically adjusting the warning level, we can ensure that appropriate measures are taken when necessary while reducing unnecessary waste of resources. For example, we can increase the intensity and frequency of monitoring when deformation accelerates, while we can relax our vigilance when deformation stabilizes or slows down, so as to allocate human, material and technical resources more effectively.

[0061] Properly set warning thresholds can help reduce false positives (i.e., alarms are triggered when there is no actual risk) and false negatives (i.e., risks exist but warnings are not given in time). This not only improves the reliability of the system, but also enhances users’ trust in the system.

[0062] This method can be flexibly adjusted according to the actual conditions in different regions and geological conditions. For example, in some areas prone to rapid deformation, a lower δ value can be set to maintain a high warning sensitivity; while in relatively stable areas, a larger δ value can be used to reduce overreaction.

[0063] The adjustment strategy based on mathematical formulas and statistical models can be easily programmed and run automatically without human intervention, which is particularly important for large-scale long-term monitoring projects. As more data is accumulated, the selection of adjustment coefficient δ and other parameters can be continuously optimized to further improve the performance and accuracy of the system.

[0064] Update of the contingency plan: Level 1 warning: arrange daily inspections, increase the frequency of data recording, and notify on-site staff on duty to pay attention.

[0065] Level 2 warning: Suspend small-scale construction activities in the surrounding area, organize technical personnel to conduct detailed inspections of key areas, and prepare emergency supplies.

[0066] Transition between different levels: When the warning level is raised or lowered, there should be a smooth transition mechanism to ensure that response measures are switched in an orderly manner to avoid waste of resources or insufficient response.

[0067] In the model training and operation process, the dynamic threshold adjustment module is embedded after the prediction and warning stages and before the execution of the disposal plan, so that the early warning system can adapt to changes in the formation, continuously optimize the accuracy of the early warning, and ensure project safety.

[0068] The present invention can ensure that there are targeted response measures at different warning levels through clear disposal plans, so that relevant personnel can take quick action and improve the emergency response speed. According to the different warning levels, resources are reasonably arranged to avoid unnecessary waste of resources. For example, in the case of a first-level warning, it is only necessary to strengthen inspections and data recording, while in the case of a second-level warning, it is necessary to suspend construction and prepare emergency supplies. Clear warning level division and corresponding disposal plans help protect the safety of on-site workers, especially in the case of a higher-level warning, timely preventive measures can reduce potential risks.

[0069] By embedding the dynamic threshold adjustment module into the prediction and warning stages and before the execution of the disposal plan, the early warning system can adapt to changes in the strata and continuously optimize the accuracy of early warnings, thereby ensuring project safety and promoting the long-term stable operation of the system.

[0070] When the warning level changes, the smooth transition mechanism ensures the orderly switching of response measures, which not only prevents the waste of resources due to sudden upgrades or downgrades, but also avoids the problem of insufficient response, and ensures continuity and stability.

[0071] There are several ways to transition between different levels: The first one: Phased implementation strategy When the warning level is upgraded from a lower level (such as level 1) to a higher level (such as level 2), instead of immediately implementing all the response measures of the higher level, the response intensity is gradually increased in stages. For example, the inspection frequency and scope are increased first, and then it is decided whether further suspension of construction activities is needed according to the situation.

[0072] Lowering of warning level: Similarly, when the level drops from a higher level to a lower level, all prevention and control measures should not be relaxed immediately, but the monitoring intensity should be gradually reduced and normal operations should be resumed. For example, a higher data recording frequency should be maintained for a period of time, and then restored to normal level after confirming that there are no abnormalities.

[0073] Second type: Establish an "early warning buffer zone", that is, set an observation period before officially entering a new warning level. During this period, although the new warning standards have been met, they are not fully implemented in accordance with the new level standards, but some intermediate measures are taken. It can be evaluated whether it is really necessary to raise the warning level and provide more information for the final decision.

[0074] Specifically: set a buffer threshold between the current warning level and the next warning level. For example, if the error threshold of the first warning is e 1 , the error threshold of the second-level warning is e 2 , then you can 1 and e 2 Select an appropriate value as the upper and lower limits of the buffer; The reasonable length of the observation period is determined according to the specific application scenario. The observation period depends on the speed and urgency of the formation deformation and can be several days or several hours.

[0075] During the buffer period, continue to closely monitor the development of stratum deformation, record relevant data, and analyze these data to determine whether there is indeed a trend of accelerated deformation or whether deformation has begun to slow down.

[0076] During the buffer period, some milder response measures can be implemented, such as increasing the frequency of inspections and strengthening data analysis, without having to take more stringent actions immediately (such as suspending construction).

[0077] Automated tools and model predictions can also be used to help decision makers make more informed choices during the buffer period. This may include using machine learning algorithms to predict future deformation trends or referencing best practices from similar historical cases.

[0078] In e 1 and e 2 An appropriate value is selected as the upper and lower limits of the buffer. In this embodiment, the upper and lower limits of the buffer can be defined by using the mean and standard deviation of the error sequence: Set the lower limit of the buffer Add a small safety margin to the current warning level threshold: ; in, The ratio is preset as a safety margin.

[0079] Set buffer limit Subtract another small safety margin for the next warning level threshold: ; in, is the warning threshold of the next level, It is also the preset ratio for safety margin. and All are positive numbers.

[0080] The present invention adjusts the parameters and (i.e. the ratio of safety margin) can flexibly respond to different application scenarios and risk preferences. For example, in projects with extremely high safety requirements, a larger and Increase the buffer width; if resources are limited, a smaller buffer can be used and .

[0081] A properly set safety margin can help filter out the impact of short-term fluctuations or outliers, avoid false positives due to accidental factors, and reduce underreporting caused by ignoring real risks.

[0082] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art can understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0083] See also Figure 3,The second embodiment of the present invention proposes a stratum deformation early warning device based on multi-parameter dynamic monitoring, based on a stratum deformation early warning method based on multi-parameter dynamic monitoring, the device comprises: a fixed pile 1, a groundwater level monitoring component, a pore water pressure monitoring component, a soil pressure monitoring component and a ground settlement monitoring component; The fixing pile 1 is fixedly implanted in the foundation and the top end thereof protrudes from the ground to a set height, and is used to support each component; The underground water level monitoring assembly includes a water level sensor 8 buried in the foundation and arranged in the water level layer 11 for monitoring the change of the underground water level; The pore water pressure monitoring assembly includes a pore water pressure gauge 6 buried in soil layers at different depths, which is used to measure the pore water pressure in each soil layer in real time; The soil pressure monitoring component includes a soil pressure sensor buried in the rock and soil layer 12, which is used to measure the soil pressure in real time; The ground settlement monitoring assembly includes a reference plate 2 fixed on the ground and a displacement sensor 7. The displacement sensor 7 is connected to the reference plate 2 and is arranged under the reference plate 2 and the fixed pile 1 to monitor the displacement of the reference plate 2 and obtain stratum deformation data.

[0084] The device also includes a data acquisition and storage unit 5 and a data transmission and early warning unit 4. The data acquisition and storage unit 5 is electrically connected to a water level sensor 8, a pore water pressure gauge 6, a soil pressure sensor and a displacement sensor 7, and is used to obtain pore water pressure, groundwater level changes, soil pressure and ground settlement data, and transmit them to a remote monitoring center through the data transmission and early warning unit 4.

[0085] In this embodiment, the data acquisition and storage unit 5 is debugged, and the data acquisition frequency, transmission parameters and storage parameters are set to ensure that the monitoring data can be stably collected, processed, stored and transmitted to the remote monitoring center.

[0086] The present invention performs drilling operations in the selected monitoring area during use, and determines the burial depth of the groundwater level monitoring pipe 9 and the pore water pressure gauge 6 according to geological survey data. The drilling depth needs to penetrate the soil layer that may affect the monitoring result to reach the stable stratum.

[0087] Place the groundwater level monitoring pipe 9 into the borehole, make the permeable hole 10 within the expected groundwater level change range, fill fine sand around the pipe, install the water level sensor 8, and connect the data line to the data acquisition and storage unit 5. At the same time, install a sealing cover on the top of the groundwater level monitoring pipe 9, and record the buried depth of the pipe body.

[0088] According to the design requirements, the pore water pressure gauge 6 is buried in the soil layer at different depths so as to be in close contact with the surrounding soil. The waterproof and corrosion-resistant data line is connected to the data acquisition and storage unit 5, and the data line is fixed and protected to prevent it from being damaged.

[0089] The fixed pile 1 is implanted in the center of the drill hole and fixed firmly by grouting or other means to ensure its verticality and stability. The top of the fixed pile 1 protrudes a certain height from the ground. The surface of the fixed pile 1 is inspected and maintained to ensure that it is free of damage and corrosion.

[0090] The reference plate 2 is installed on the ground to make it fit closely with the ground and perform anti-skid treatment. The displacement sensor 7 is installed below the reference plate 2 and the fixed pile 1 , the sensor position is adjusted, and the data line is connected to the data acquisition and storage unit 5 .

[0091] The data acquisition and storage unit 5 collects groundwater level, pore water pressure and ground settlement data at a set frequency and transmits them to the remote monitoring center in real time. The staff analyzes and processes the data at the remote monitoring center. When an abnormal change in the data is found and an early warning is received, on-site inspections are carried out in a timely manner or corresponding engineering measures are taken, such as adjusting groundwater level control measures, reinforcing the foundation, etc. The bottom of the fixed pile 1 adopts a conical design to facilitate implantation in the foundation, and grouting and other methods can be used during the implantation of the fixed pile 1 to further enhance its stability.

[0092] Among them, the pore water pressure gauge 6, soil pressure sensor, and pore water pressure gauge 6 collect data every 15 minutes, and the displacement sensor 7 records data every 30 minutes. The collected data is attached with a timestamp and transmitted to the data processing center in real time through the data transmission and early warning unit 4 by wired or wireless transmission.

[0093] The data line of the pore water pressure meter 6 adopts a waterproof and corrosion-resistant cable to ensure the stability and reliability of data transmission.

[0094] The reference plate 2 of the ground subsidence monitoring assembly is made of high-strength aluminum alloy material, which has the characteristics of light weight, high strength and good corrosion resistance, and its surface is anti-slip treated to ensure stability on the ground.

[0095] The data acquisition and storage unit 5 adopts a modular design, which is easy to install and maintain. A data storage module is arranged inside the unit, which can temporarily store monitoring data when the network signal is poor, and automatically upload the data to the remote monitoring center after the signal is restored.

[0096] As a further explanation of the present invention, the groundwater level monitoring component also includes a groundwater level monitoring pipe 9, a water permeable hole 10 is opened on the pipe wall of the groundwater level monitoring pipe 9, and a filter is installed on the water permeable hole 10 to prevent impurities from entering the pipe and affecting the water level measurement. A water level sensor 8 is installed inside the groundwater level monitoring pipe 9.

[0097] A sealing cover is provided on the top of the groundwater level monitoring pipe 9 to prevent debris from entering the pipe and affecting the measurement accuracy. At the same time, scale markings are provided on the pipe body to facilitate determination of its buried depth during installation and maintenance.

[0098] The underground water level monitoring pipe 9 in this embodiment is made of corrosion-resistant PVC material, and its water permeable holes 10 are evenly distributed in the lower part of the pipe body. The filter screen is made of stainless steel, which can effectively filter impurities such as mud and sand to ensure the accuracy of water level measurement.

[0099] The pore water pressure meter 6 adopts a high-precision vibrating-wire pore water pressure meter 6, which can accurately measure tiny changes in pore water pressure and has good long-term stability.

[0100] The displacement sensor 7 is a laser displacement sensor 7 , which has the advantages of high precision, high sensitivity and non-contact measurement, and can quickly and accurately measure the displacement between the reference plate 2 and the fixed pile 1 .

[0101] The fixing pile 1 is made of high-strength alloy steel material, has good stability and corrosion resistance, and is provided with an anti-rust coating on its surface to extend its service life.

[0102] The device also includes a power supply module 3, which includes a solar panel and a lithium battery pack. The solar panel is used to charge the lithium battery pack. The lithium battery pack and the solar panel are electrically connected to the water level sensor 8, the pore water pressure gauge 6, the soil pressure sensor and the displacement sensor 7 and provide power.

[0103] Specifically, the solar panels charge the lithium batteries during the day when there is sufficient sunlight, while maintaining the operation of the device. The lithium battery pack serves as a backup power source, continuously supplying power to the entire monitoring device when there is insufficient sunlight at night or on rainy days, ensuring uninterrupted operation of the monitoring operation.

[0104] The third embodiment of the present invention proposes a stratum deformation early warning system based on multi-parameter dynamic monitoring, based on a stratum deformation early warning method based on multi-parameter dynamic monitoring, the system comprises: A data set construction module configured to obtain multiple parameters for monitoring stratum deformation and time series data of stratum deformation data as original data sets, wherein the multiple parameters at least include pore water pressure, groundwater level change and soil pressure; A preprocessing module, configured to remove outliers in the original data set and perform normalization processing; A weighting module, configured to respectively calculate the standard deviation of multiple parameters and the correlation coefficient with the formation deformation, and calculate the amount of information between the standard deviation and the correlation coefficient of each parameter, calculate the weight values ​​of the multiple parameters based on the amount of information, weight the normalized multiple parameters according to the weight values, obtain the input feature vector after weighting, and construct a data set to be trained; A training module, configured to input the data set to be trained into a pre-built stratum deformation prediction model for training; wherein the stratum deformation prediction model is built based on a GAN-LSTM network; A prediction module is configured to collect pore water pressure, groundwater level change and soil pressure of the formation in real time according to a set collection frequency as input data, and input them into a trained formation deformation prediction model to obtain formation deformation prediction data; The early warning module is configured to calculate the absolute error between the formation deformation prediction data and the historical formation deformation data in real time and form an error sequence, calculate the early warning threshold based on the error sequence, adjust the early warning threshold by the deformation trend of the formation deformation prediction data within a set time, divide the early warning level according to the adjusted early warning threshold, and execute the corresponding disposal plan according to the early warning level corresponding to the formation deformation prediction data.

[0105] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0106] It should be noted that the above embodiment provides a stratum deformation early warning system based on multi-parameter dynamic monitoring, which is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of the present invention can be decomposed or combined. For example, the modules in the above embodiments can be combined into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps, and are not regarded as improper limitations of the present invention.

[0107] An electronic device according to a fourth embodiment of the present invention includes: at least one processor; and a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the above-mentioned formation deformation early warning method based on multi-parameter dynamic monitoring.

[0108] A fifth embodiment of the present invention is a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to be executed by the computer to implement the above-mentioned formation deformation early warning method based on multi-parameter dynamic monitoring.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process and related instructions of the storage device and processing device described above can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0110] Those skilled in the art should be able to appreciate that the modules and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two, and the programs corresponding to the software modules and method steps can be placed in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field. In order to clearly illustrate the interchangeability of electronic hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0111] In the description of the present invention, the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings, which are only for the convenience of description, and do not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0112] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0113] The term "comprise" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus / device.

[0114] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A stratum deformation early warning method based on multi-parameter dynamic monitoring, characterized in that: The method includes: Acquire multiple parameters for monitoring stratum deformation and time series data of stratum deformation data as original data sets, wherein the multiple parameters at least include pore water pressure, groundwater level change and soil pressure; Eliminate outliers in the original data set and perform normalization processing; Calculating the standard deviation of multiple parameters and the correlation coefficient with formation deformation respectively, and calculating the information amount between the standard deviation and the correlation coefficient of each parameter, calculating the weight values ​​of multiple parameters based on the information amount, weighting the normalized multiple parameters according to the weight values, obtaining the input feature vector after weighting, and constructing the data set to be trained; Inputting the data set to be trained into a pre-built stratum deformation prediction model for training; wherein the stratum deformation prediction model is built based on a GAN-LSTM network; The pore water pressure, groundwater level change and soil pressure of the formation are collected in real time according to the set collection frequency as input data, and input into the trained formation deformation prediction model to obtain formation deformation prediction data; The absolute error between the formation deformation prediction data and the historical formation deformation data is calculated in real time to form an error sequence. The warning threshold is calculated based on the error sequence. The warning threshold is adjusted according to the deformation trend of the formation deformation prediction data within a set time. The warning level is divided according to the adjusted warning threshold. According to the warning level corresponding to the formation deformation prediction data, the corresponding disposal plan is executed.

2. The method for early warning of stratum deformation based on multi-parameter dynamic monitoring according to claim 1 is characterized in that: The data set to be trained is input into the pre-built formation deformation prediction model for training, and the training method is as follows: The data set to be trained is used as input data and is sequentially input into the formation deformation prediction model according to time steps. The input data is analyzed in time series through the LSTM network. Use a fully connected layer to extract features from the hidden state of the last time step of the LSTM network to predict the formation deformation value; The mean square error is used as the loss function to adjust the parameters of the LSTM network, and the loss function is minimized by the Adam optimization algorithm; The trained LSTM network is used as the generator of the GAN network to generate formation deformation prediction data; Construct a multi-layer perceptron as the discriminator of the GAN network to distinguish the real stratum deformation data from the stratum deformation prediction data output by the generator; Adjust the parameters of the generator and discriminator based on a preset loss function; The Adam optimizer is used alternately to update the generator and discriminator parameters, and a set number of rounds of iterative training are performed until convergence.

3. The method for early warning of stratum deformation based on multi-parameter dynamic monitoring according to claim 1 is characterized in that: The method for calculating the warning threshold based on the error sequence is as follows: ； in, is the first level warning threshold, is the second level warning threshold, is the mean of the error sequence, is the standard deviation of the error series, , is the preset empirical coefficient.

4. The method for early warning of stratum deformation based on multi-parameter dynamic monitoring according to claim 1 is characterized in that: The warning threshold is adjusted by the deformation trend of the formation deformation prediction data within a set time, and the method is as follows: Based on the sliding window algorithm, the past k Linear fitting is performed on the formation deformation prediction data of each time step and the slope is obtained as the deformation trend; If the slope is positive and greater than a preset threshold, lower the warning threshold; If the slope is negative, the warning threshold is increased.

5. The method for early warning of stratum deformation based on multi-parameter dynamic monitoring according to claim 4 is characterized in that: The method of lowering the warning threshold and raising the warning threshold is as follows: Lower the warning threshold: ; Increase the warning threshold: ; in, To lower or raise the warning threshold, i is 1 or 2, is the original warning threshold, is the adjustment factor.

6. The method for early warning of stratum deformation based on multi-parameter dynamic monitoring according to claim 5 is characterized in that: The method for dividing the warning levels according to the adjusted warning thresholds is as follows: When the error When there is no warning; when It is a Level 1 warning when It is a level 2 warning.

7. A stratum deformation early warning device based on multi-parameter dynamic monitoring, based on a stratum deformation early warning method based on multi-parameter dynamic monitoring according to any one of claims 1 to 6, characterized in that: The device comprises: a fixed pile (1), a groundwater level monitoring component, a pore water pressure monitoring component, a soil pressure monitoring component and a ground settlement monitoring component; The fixing pile (1) is fixedly implanted in the foundation and has its top end protruding from the ground to a set height, and is used to support each component; The underground water level monitoring component comprises a water level sensor (8) buried in the foundation and arranged in the water level layer (11) for monitoring changes in the underground water level; The pore water pressure monitoring assembly comprises a pore water pressure gauge (6) buried in soil layers at different depths, and used for measuring the pore water pressure in each soil layer in real time; The soil pressure monitoring component comprises a soil pressure sensor buried in the rock and soil layer (12) and used for measuring the soil pressure in real time; The ground subsidence monitoring assembly comprises a reference plate (2) fixed on the ground and a displacement sensor (7); the displacement sensor (7) is connected to the reference plate (2) and is arranged below the reference plate (2) and the fixed pile (1) to monitor the displacement of the reference plate (2) and obtain stratum deformation data.

8. The stratum deformation early warning device based on multi-parameter dynamic monitoring according to claim 7 is characterized in that: The groundwater level monitoring assembly further comprises a groundwater level monitoring pipe (9), a water permeable hole (10) being provided on the pipe wall of the groundwater level monitoring pipe (9), a filter screen being installed on the water permeable hole (10), and a water level sensor (8) being installed inside the groundwater level monitoring pipe (9).

9. The stratum deformation early warning device based on multi-parameter dynamic monitoring according to claim 7 is characterized in that: The device further comprises a data acquisition and storage unit (5) and a data transmission and early warning unit (4); the data acquisition and storage unit (5) is electrically connected to a water level sensor (8), a pore water pressure gauge (6), a soil pressure sensor and a displacement sensor (7) and is used to obtain data on pore water pressure, groundwater level change, soil pressure and ground subsidence, and transmit the data to a remote monitoring centre via the data transmission and early warning unit (4).

10. The stratum deformation early warning device based on multi-parameter dynamic monitoring according to claim 7, characterized in that: The device further comprises a power supply module (3), the power supply module (3) comprising a solar panel and a lithium battery pack, the solar panel being used to charge the lithium battery pack, the lithium battery pack and the solar panel being electrically connected to the water level sensor (8), the pore water pressure gauge (6), the soil pressure sensor and the displacement sensor (7) and providing power.

Images